Yaw-adjusted wind power curve modeling: A local regression approach

Accurate estimation of wind power curves using field data is instrumental to several wind farm operations including productivity assessment, power output estimation, operations and maintenance, among others. Existing methods for estimating wind power curves mainly rely on environmental variables (e.g., wind speed, direction, density) as inputs to construct the wind-to-power relationship. This paper attempts to integrate yaw misalignment as an additional input to power curve models, constructing what is referred to hereinafter as yaw-adjusted wind power curves.Our analysis shows that integrating yaw misalignment into power curves is non-trivial, largely due to the overwhelming impact of environmental variables (mainly wind speed) on a turbine's power output, which obscures the secondary effect of yaw errors on power production. In response, we propose a local-regression-based method which reconstructs the yaw-to-power relationship conditional on an effective neighborhood of environmental variables. Tested on operational data from two onshore wind turbines in France, our proposed approach achieves significant improvements, in terms of power estimation accuracy, relative to a set of prevalent statistical-and machine-learning-based power curve models.

Automated summary

Accurate estimation of wind power curves is crucial for various wind farm operations. Existing methods typically use environmental variables to construct the relationship between wind and power. This paper proposes a method that integrates yaw misalignment as an additional input, resulting in improved accuracy. The proposed method uses a local-regression-based approach to reconstruct the relationship between yaw and power, conditional on environmental variables. Testing on real data from two onshore wind turbines in France demonstrates significant improvements compared to existing models.